Video Transcription & Indexing Pipeline

from mixpeek import Mixpeek

client = Mixpeek(api_key="YOUR_API_KEY", namespace="transcripts")

# 1. A bucket for the recordings, and a collection that cuts each one at pauses
#    and transcribes and embeds every segment
bucket = client.buckets.create(
    bucket_name="meeting-recordings",
    bucket_schema={"properties": {"recording": {"type": "video"}}},
)
collection = client.collections.create(
    collection_name="meetings",
    source={"type": "bucket", "bucket_ids": [bucket["bucket_id"]]},
    feature_extractor={
        "feature_extractor_name": "multimodal_extractor",
        "version": "v1",
        "parameters": {
            "split_method": "silence",
            "run_transcription": True,
            "run_transcription_embedding": True,
        },
    },
)

# 2. Upload and process. The managed extractor does not label speakers, so
#    diarization runs in your own pipeline when you need it.
client.buckets.upload(
    bucket["bucket_id"],
    blobs=[{"property": "recording", "type": "video", "data": "s3://your-bucket/meeting-recordings/weekly-sync.mp4"}],
)
client.collections.trigger(collection["collection_id"])

# 3. Read the transcript one segment at a time, with its time range
docs = client.documents.list(collection["collection_id"], page_size=50)
for doc in docs["results"]:
    print(doc["start_time"], doc["end_time"], doc["transcription"])